CNN deep learning-based image to vector depiction
In the computational science and engineering domains, the depiction of picture information remains an intricate problem. Such a description needs an accurate recognition of various objects and individuals together with their attributes, correlations, and panorama information. Based on this fact, we...
Saved in:
| Main Authors: | , , , |
|---|---|
| 格式: | Article |
| 出版: |
Springer Nature
2023
|
| 主題: | |
| 在線閱讀: | http://eprints.utm.my/105917/ http://dx.doi.org/10.1007/s11042-023-14434-w |
| 標簽: |
添加標簽
沒有標簽, 成為第一個標記此記錄!
|
| 機構: | Universiti Teknologi Malaysia |
| 總結: | In the computational science and engineering domains, the depiction of picture information remains an intricate problem. Such a description needs an accurate recognition of various objects and individuals together with their attributes, correlations, and panorama information. Based on this fact, we depict the image contents in the natural language or image description generation methods using the convolutional neural networks (CNNs)-assisted deep learning (CNN-DL) approach, wherein the images are transformed to vectors. The DL and study attributes via the machine-learned data were used to construct the complete pictures from the real world. Two sections were considered based on image classification for CNN’s improvement method to develop a classification model and the good results of the classification via a novel method for describing an image to the vector of each object in the image. The learning and relationship activity included all the essential categorizing and classifying entities. In addition, the developed system was extended to handle the open detection and hazards classification. The performance evaluation (using the CIFAR dataset) of the newly developed system revealed its better strength and flexibility in managing the test images from a new-fangled and isolated field than the reported techniques. |
|---|